[j-nsp] traceroute in mpls vpn's not showing P hops

Chris Burton chris.burton at speakeasy.net
Wed Aug 23 21:38:51 EDT 2017


Given that LSP's are by there very nature unidirectional you will need a 
return LSP unless you plan export your VRF routes into your underlying 
network.  The path that that return LSP takes should not make a 
difference unless you are testing specific return path(s), so it should 
not interfere with your learning of RSVP, actually that operation is no 
different than much of standard L3 routing in the wilds of the Internet, 
you are not guaranteed the same path back as the path taken out beyond 
the first hop or few in most cases.


-C

On 08/23/2017 12:10 PM, Aaron Gould wrote:
> Thanks, opps, I forgot to tell y'all I tried "set protocols mpls
> icmp-tunneling" already and it didn't work.
>
> Also, I need to mention that the vrf between the (2) CE's is built on top of
> a unidirectional rsvp lsp.... I'm testing rsvp te to learn about it and I
> only built the lsp in one direction to influence traffic one-way... So
> Chris's comments about "sure you have LSP's in both directions"  makes me
> wonder if I do the return rsvp LSP for the underlying rsvp te would that fix
> it ? (furthermore, must the return rsvp te lsp follow the exact same path
> for traceroutes to work ?! this would seem really limiting)
>
> I mean, I have LDP running in the core also, this might be obvious since the
> vrf is functional, and I'm pretty sure I would need the mpls control plane
> to work bidirectionally to have the vrf function as it is.
>
> Let me know what y'all think.
>
> Ce---pe----p----p----p----p----p----pe----ce
>
> ---------------rsvp te from pe to pe -------->   this direction only... I
> allow the igp to decide the return path in the opposite direction
>
>
> -Aaron
>
>
>
>
>
>
>
>
>



More information about the juniper-nsp mailing list